diff --git a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md index 77009e19..52b27cce 100644 --- a/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md +++ b/.claude/knowledge/pr-x12-h268-morton-wgpu-synergies.md @@ -114,7 +114,7 @@ any float-path GPU bit-exactness claim. | WHP-1..4 (+GPU arm) | two-algebra pyramid parity | per OGAR canon | magnitude side stays CPU | | Plan E bench | bits/Gaussian on Mip-NeRF 360 | ≤4 bits | R-10 re-derived; web-streaming claim withdrawn | | a2ui N2 | wgpu `webgl` feature + texture upload, wasm32-tested | render parity headless vs browser | GPU raster tier deferred; CPU raster only | -| PROBE-SPRITE-REPLAY | x265 I/P/B grammar over HHTL-anchored splat sprites with helix motion codes (plan `x265-sprite-replay-probe-v1.md`; NOT H.268, NOT x265 bit-parity) | replay determinism CPU==wasm on sprite states; B-frame bidirectional consistency; helix quantization bound reported | object-level helix motion collapses back into a dense per-splat MV field — the sprite amortization dies as stated | +| PROBE-SPRITE-REPLAY | x265 I/P/B grammar over HHTL-anchored splat sprites with helix motion codes (plan `x265-sprite-replay-probe-v1.md`; NOT H.268, NOT x265 bit-parity) | replay determinism CPU==wasm on sprite states; B-frame bidirectional consistency; helix quantization bound reported | object-level helix motion collapses back into a dense per-splat MV field — the sprite amortization dies as stated → **VERDICT (wave 2): PASS-AT-SIGNED360 scoped / ResidueEdge-24bit insufficient; see wave-2 sub-table** | **Run this wave — h268-probe-wave-v1 (2026-07-16, reviewer-adjudicated; probes live in lance-graph `bgz-tensor`/`jc`/`helix`):** @@ -125,6 +125,14 @@ probes live in lance-graph `bgz-tensor`/`jc`/`helix`):** | PROBE-SIG-CHECKSUM | PASS (depth-2 blind-spot bound) | replay 0.0; tree-like 0.0<1e-9; non-tree 11.31>0.05 | §10 [S] leg confirmed; digest ≠ catch-all — parallel-chord displacement is signature-invisible at depth 2 | | PROBE-WALK-SPECTRUM | KILL of §10(g) "decorrelated" half | period-17 \|R\| 0.998/0.996/0.994; lattice max 0.875 vs PRNG 0.0205 = 42.7× | §10(g) corrected: known structure ≠ low correlation; D-QUANTGATE / bijectivity unaffected | +**Wave 2 — x265-sprite-replay + WH-MAG-2 (2026-07-16, reviewer-adjudicated; +probes in lance-graph `helix`/`bgz-tensor`):** + +| Probe | Verdict | Numbers | Consequence | +|---|---|---|---| +| PROBE-SPRITE-REPLAY | PASS-AT-SIGNED360 (scoped) / ResidueEdge-24bit INSUFFICIENT; KILL did not fire | Signed360 mean/max 0.0; ResidueEdge mean 9.98 / max 42.4 (Pos 3.55, Neg 16.4); B bidir-delta 6.04 | One Signed360 (48-bit) code per sprite per P-frame reconstructs motion + B-frames exactly — but for helix-MANIFOLD motion (ground truth drawn from helix's own lift); proves capacity+round-trip+sign, not arbitrary-motion generality ([H], independent-GT probe named). ResidueEdge-24bit unusable: hemisphere-blind + 8-bit-rim rank-adjacency hazard → Signed360-only motion primitive | +| PROBE-WH-MAG-2 | UPGRADES the §10 WH-magnitude [H] leg: bare NOT-TRANSFERRING → pairing TRANSFERS on structured tiles | gradient+spike escB/A 0.815 + cenB/A 0.209 (both <0.9); heavy-tailed 0.998 / 1.232; noise 1.66 / 1.66; bareB/A 0.929/1.317/1.869 reproduced; esc_frac 0.098 | WH transfers ONLY paired with passthrough-escape + centroid-residual (the shipped row codec E-PALETTE-RESIDUAL-LADDER-1) and ONLY on structured tiles; NOT bare, NOT heavy-tailed (escape masks to parity; centroid hurts), NOT noise. escB/cenB are compound (no direct+escape/direct+centroid control) — transfer is a codec-pairing property, WH's isolated margin unmeasured; centroid is class-dependent | + ## 7. Comma closure — the replayable irrational (constants correction folded in) The Pythagorean comma is the residue of a stack of pure fifths that never @@ -458,6 +466,21 @@ fourth-mode/anchor discussion (§8): FailureTicket). PROBE-WH-MAG-2's natural home is therefore WH re-evaluated UNDER the mode grammar (heavy-tailed tiles routed to Escape before the cascade), not a wait for new machinery. + **RESULT-2 (2026-07-16, PROBE-WH-MAG-2, upgrades this leg):** run under + the escape+centroid pairing (the deferral-condition home named above), + WH TRANSFERS on structured tiles — gradient+spike escB/A **0.815** + + cenB/A **0.209** both clear the <0.9 bar — but NOT bare (0.929, leg-1 + reconfirmed), NOT heavy-tailed (escB/A 0.998 = escape masks WH to + parity; cenB/A 1.232 = centroid hurts zero-mean+outliers), NOT noise + (1.66/1.66). The [H] leg is upgraded from "bare NOT-TRANSFERRING" to + "TRANSFERS with the shipped WH+escape+centroid pairing on STRUCTURED + tiles only." Caveat: escB/cenB are compound (Path A has no + escape/centroid sibling), so the win is a codec-pairing property — WH's + isolated marginal value beyond escape+centroid is unmeasured; bareB/A + stays the only clean WH isolation and stays NEUTRAL. Cross-ref + lance-graph E-X265-PROBE-WAVE-2-RESULTS, + `crates/bgz-tensor/src/adaptive_codec.rs` mod + probe_wh_mag::probe_wh_mag_2. - **[S] Signature as the replayable-trajectory checksum.** The x264 contrast in §7 asked for a replayable, checksummable stream; the Hambly–Lyons signature is the canonical path digest with a uniqueness diff --git a/.claude/plans/x265-sprite-replay-probe-v1.md b/.claude/plans/x265-sprite-replay-probe-v1.md index 7f24ce5a..529ced17 100644 --- a/.claude/plans/x265-sprite-replay-probe-v1.md +++ b/.claude/plans/x265-sprite-replay-probe-v1.md @@ -93,3 +93,43 @@ CPU/wasm decode, harness), filigree adjudication vs the pass/KILL bands, central gates, PR, autonomous merge (standing authority). The wgpu tier may land as a second commit gated on the shared harness; if it slips, the CPU/wasm probe stands alone. + +## Results (2026-07-16 — PROBE-SPRITE-REPLAY-CORE, reviewer-adjudicated) + +**Verdict: PASS-AT-SIGNED360 (scoped) / ResidueEdge-24bit INSUFFICIENT. The KILL band did NOT fire.** + +Probe: `lance-graph crates/helix/src/sprite_replay.rs` (`probe_sprite_replay_core`), +CPU-native core only (splat3d EWA raster + wgpu tiers deferred per the scope +guard); N=8 sprites, TOTAL=240, 6 anchors (I+5P), GOP `I B B P ...`. Deterministic +(SplitMix64); the test asserts determinism + finiteness + counts, never a verdict. + +| Metric (world-unit Euclidean position error) | ResidueEdge (24-bit) | Signed360 (48-bit) | +|---|---|---| +| I+P anchor motion fidelity — mean / max | 9.98 / 42.43 | **0.0 / 0.0** | +| Sign::Pos sprites — mean / max | 3.55 / 17.60 | 0.0 / 0.0 | +| Sign::Neg sprites — mean / max | 16.41 / 42.43 | 0.0 / 0.0 | +| B-frame bidir-delta — mean / max | 6.04 / 22.50 | 0.0 / 0.0 | +| B-frame vs-truth — mean / max | 10.45 / 41.57 | 0.0 / 0.0 | + +- **KILL did not fire** (plan §4): object-level helix motion does NOT need a + dense per-splat residual field — one `Signed360` code per sprite per P-frame + reconstructs motion + parametric B-frames exactly. +- **PASS is scoped (the load-bearing caveat).** Ground-truth motion is drawn + from helix's OWN `signed_lift` template, so the 0.0 Signed360 result proves + CAPACITY (16-bit azimuth indexes n∈0..240 trivially) + ROUND-TRIP (the + hand-built public-primitive inverse recovers n and sign exactly) + SIGN + carriage — NOT that arbitrary independently-captured object motion fits the + helix manifold. "P-frame = one Signed360 code" is FEASIBLE at 48-bit for + **helix-manifold** motion [proven]; arbitrary-motion generality stays [H], + gated on a follow-up probe with independent ground truth. +- **ResidueEdge (24-bit) INSUFFICIENT — Signed360-only ruling.** Two + independent failures: (1) hemisphere-blind (no sign bit → Neg sprites mirror + wrong: mean 16.4 / max 42.4); (2) rank-adjacency hazard (8-bit `end_idx` + bucket; a ±1 rank error under golden-spiral ~137.5° spacing lands a distant + point — even Pos sprites: mean 3.55 / max 17.6). The sprite motion primitive + should be Signed360; ResidueEdge is a residue-edge codec, not a motion code. + +**Follow-up (named, deferred):** an arbitrary-independent-ground-truth motion +probe (a captured 3-D trajectory the encoder did not generate) to promote the +helix-manifold [proven] result toward arbitrary-motion [H]. The wgpu/wasm decode +tiers (plan §Decode tiers b/c) remain deferred on the shared PROBE-GPU-LUT harness.